1. Field of the Invention
The present invention relates to a configuration with an electrically operated appliance with a voice input unit and with a voice processing unit that derives necessary control signals from spoken input instructions for controlling functions of the appliance, and a voice input method with correction of interference signals.
International Publication WO 98/55992, corresponding to U.S. Pat. No. 6,230,137 to Has et al., discloses a configuration and a method in which a household appliance is described that can be controlled by voice signals. To avoid malfunctions, for example, due to voice signals that do not originate from an operator but rather, for example, from a radio, the publication describes configuring a control program such that successive voice signals lead to the formation of a control instruction only if the successive voice signals are input within a predefinable time.
Furthermore, a product with the designation xe2x80x9cSICARE pilotxe2x80x9d (from evosoft Software-Vertrieb [sales] GmbH, D-90411 Nuremberg/Germany) is already in the REHA field and includes a mobile appliance that converts spoken instructions into signals. The signals are output by an infrared transmitter, for example, for controlling television appliances, video recorders, hi-fi equipment, telephone and lighting, by a radio transmitter, for example, for controlling house doors, wheelchairs, personal paging systems, and alarm systems or by wire-bound interfaces, for example, for controlling electrically driven wheelchairs and beds that can be adjusted by motor. The prior art product, which is intended for disabled people, permits a maximum of 64 instructions to be stored. The mobile appliance transmits the instruction for the desired action, for example, xe2x80x9ctelevision on,xe2x80x9d to the respective appliance automatically through the previously selected interface. Each individual instruction can address a plurality of appliances simultaneously. The prior art product is trained to the individual voice pattern of a person and, thus, carries out speech recognition on a speaker-dependent basis. In an ideal case, other persons are not able to form signals that control the appliance from spoken instructions.
Such voice recognition is implemented in the prior art product such that predefined words (for example xe2x80x9chouse,xe2x80x9d xe2x80x9clight,xe2x80x9d xe2x80x9cfour,xe2x80x9d or xe2x80x9cminusxe2x80x9d) can be spoken in accordance with a menu tree with a plurality of branches, it being possible to speak the individual words at any desired intervals. If the voice input is prematurely aborted as a result of an indisposition of the disabled person, individual words, for example, from a radio or television, which are provided in the branch of the menu tree after the point at which the voice input was aborted may lead to the formation of signals that control one or more appliances.
Furthermore, German Published, Non-Prosecuted Patent Application DE 196 37 352 A1 discloses a method for automatic voice recognition, reference utterances being firstly recorded for different words and then sequences of chronologically successive reference feature vectors being formed from them, which vectors are each mapped onto a uniform chronological interval. A single model of the reference utterances, whose components are composed of compensation functions, is formed for each word from the sequences of model vectors that are produced here for the reference utterances. A word that is to be recognized is processed to form a sequence of feature vectors that are each mapped onto the same chronological interval. The resulting sequence of mapping vectors is compared with the stored models in comparison steps.
Furthermore, the manual xe2x80x9cSprachsynthese und Spracherkennung [speech synthesis and speech recognition]xe2x80x9d, J. N. Holmes, R. Oldenbourg-Verlag, Munich, Vienna 1991, generally discloses that some information items relating to the phonetic content of the speech signal can be derived from the chronological profile of the speech signal. However, experience has shown that the time signal is not suitable for representing the properties of spoken language that are most important for general voice quality and perception of phonetic details. In view of the significance of the resonances and their chronological changes during verbal communication, a method for representing these features is necessary. The short-term spectrum of the signal that corresponds to the absolute value of the Fourier transformation of the time signal after it has been multiplied by a time window function with an appropriate duration may, of course, not contain any information that is not also contained in the original signal. However, it is also significantly more suitable for representation of the resonances and, in particular, for the computer-supported signal processing.
It is accordingly an object of the invention to provide an electrical appliance with voice input unit and voice input method that overcome the hereinafore-mentioned disadvantages of the heretofore-known devices and methods of this general type and that improves speech recognition despite the presence of interference noises.
With the foregoing and other objects in view, there is provided, in accordance with the invention, a configuration having an electrically operated household appliance, a voice input unit, and a voice processing unit deriving control signals from spoken input instructions for controlling functions of the household appliance, including an operating status detection unit to be connected to the household appliance and the voice processing unit, the operating status detection unit adapted to detect an operating status of at least one of the household appliance and other noise sources, to signal the operating status of the at least one of the household appliance and other noise sources to the voice processing unit, and to control the voice processing unit to perform an interference noise correction only if at least one of the noise sources is on.
With the objects of the invention in view, there is also provided a method for inputting voice signals with correction of interference signals, including the steps of interrogating an operating status of at least one noise source interfering with voice input during voice input for controlling a household appliance and performing an interference noise correction with a voice processing unit only if a noise source is on.
The invention provides an operating status detection unit that detects the operating status of the household appliance or of other noise sources and signals it to the voice processing unit, and that the voice processing unit performs an interference noise correction only if a noise source is switched on. The method according to the invention for voice input with correction of interference signals is characterized in that the operating status of at least one noise source that interferes with the voice input is interrogated during the voice input for controlling an appliance, and the voice processing unit performs an interference noise correction only if a noise source is switched on. According to the invention, therefore, when a voice signal for detection is submitted to the voice processing unit, an interference noise correction is not attempted in every case. Such a process leads to an improvement in the voice recognition in all the cases in which the voice signal is not at all subjected to interference noises. This is because the attempt to remove a nonexistent interference noise from the voice signal, or to include it in a correcting fashion, reduces the quality of the voice signal. This causes the recognition rate of the voice processing unit or voice recognition unit in the prior art to drop.
In accordance with another feature of the invention, the voice processing unit includes the interference signal of the noise source in a correcting fashion in the processing of the spoken input instructions, as a function of the respective operating status of the noise source.
For such a purpose, in accordance with a further feature of the invention, the configuration has a storage unit that is connected to the voice processing unit, and a data record that corresponds to the interference signal is stored in the storage unit for each operating status for each noise source. In a learning phase, which is carried out by the manufacturer in the laboratory or only by the end customer on the complete system that is installed in the house or in the kitchen, interference noise sources are operated in various operating modes. The respective interference noise is processed by the voice processing unit and, for example, a corresponding interference noise spectrum is stored in the storage unit as an operating-status-specific interference signal. During the actual operation of an electrical appliance according to the invention, in particular, of a household appliance in the kitchen, it is initially checked, during the voice input for controlling the appliance, whether or not interference sources analyzed in the learning phase are operating. If the sources are not operating, the voice instruction that is input is processed without noise source correction. However, if an interfering appliance is active, its operating status is detected and an interference noise data record that corresponds to the operating status is included in the signal processing or voice recognition in a correcting fashion. The voice signal that is corrected in terms of the interference noise is processed according to the prior art. The invention makes use of the fact that the number of interference sources and their typical interference noises are both limited and known in some application cases. An interference noise correction, therefore, does not need to be carried out with general correction approaches but rather it is possible for the known interference noises to be included in a correcting fashion as an alternative or an addition.
To increase the variability and application capability of the configuration according to the invention, in accordance with an added feature of the invention, on one hand, the operating status detection unit is connected to the noise source through an electrical line or through a wire free communications line, and, on the other hand, the configuration has a plurality of voice input units that are disposed spatially distributed, for example, in the kitchen or in a house.
Particularly reliable voice recognition can be implemented if the appliances that are controlled by voice input are at least a cooking surface and an extractor hood disposed over it, and the extractor hood with its various power settings is included as an interference source.
In accordance with an additional feature of the invention, the voice input unit, the voice processing unit, and the operating status detection unit are disposed in an appliance that is controlled by voice input, in particular, in the extractor hood. Under favorable thermal conditions, there is sufficient space in the extractor hood to hold the aforesaid components. Furthermore, a microphone can be disposed approximately at the level of the head of an operating person standing in front of the cooking surface that is disposed underneath the extractor hood. The microphone can have a corresponding directional characteristic to ensure that it is not subject to too high a level of interference from the extractor hood.
In accordance with yet another feature of the invention, the operating status detection unit is adapted to control the voice processing unit to perform an interference noise correction only if at least one of the noise sources is switched on.
In accordance with yet a further mode of the invention, the interference noise correction step is performed with the voice processing unit only if a noise source is switched on.
In accordance with yet an added mode of the invention, each operating status of the at least one noise source is assigned an interference noise stored as data material, the interference noise data material being included in a correcting fashion in processing of a spoken input instruction, dependent upon a currently present operating status of the at least one noise source.
In accordance with a concomitant mode of the invention, interference noises are stored as data material, each operating status of the at least one noise source is assigned a respective one of the interference noises, and a spoken input instruction is processed by correcting the spoken input instruction with the respective interference noise data material dependent upon a currently present operating status of the at least one noise source.
Other features that are considered as characteristic for the invention are set forth in the appended claims.
Although the invention is illustrated and described herein as embodied in an electrical appliance with voice input unit and voice input method, it is, nevertheless, not intended to be limited to the details shown because various modifications and structural changes may be made therein without departing from the spirit of the invention and within the scope and range of equivalents of the claims.
The construction and method of operation of the invention, however, together with additional objects and advantages thereof, will be best understood from the following description of specific embodiments when read in connection with the accompanying drawings.